Date:         Fri, 1 Mar 1996 16:03:28 GMT
Reply-To:     "Paul A. Thompson" <pat@po.CWRU.Edu>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         "Paul A. Thompson" <pat@PO.CWRU.EDU>
Organization: Case Western Reserve University, Cleveland, OH (USA)
Subject:      Re: interaction: True or False?

In a previous article, Haiyi.Xie@DARTMOUTH.EDU (Haiyi Xie) says:
 
>A friend of mine had the following questions for repeated measures analysis,
>say, one between factor such as GROUP and one within factor such as TIME:
>
>===========
>   For a while now, I have been claiming that it is more difficult to find a
>significant Group X Time interaction when there is a strong main effect for
>time, than when there is no main effect.  The intuition is that when slopes
are
 
This is simply not true.  The three terms are STATISTICALLY INDEPENDENT.
That is, there is no relationship between them.  They are independent
partitions
of some overall SS.  If you are doing the thing in multivariate mode,
and there are three time points, we work with CBU to test hypotheses.
U is test of multivariate (or within S), C between groups and both together
do combination.  For Time, we use C = I and U=1 -1 0,1 0 -1 (one of many).
for Interaction, use C = 1 -1 and U as above.
 
--
Paul Thompson, Ph.D.     |  This is a Rorshach .sig file.  Stare hard at
Department of Psychiatry |  the following terms, and tell me what you see:
Case Western Reserve Univ|  KJA:HJSDiJL*T**&h*&*e&kjrlkJeLKJ,L**M&&&r*&*.
Cleveland, OH  44106     |  JHJHGJSHrDHJeSHHeWUYnUYUjYWJeHJHaMNMnXJJsYJY!